Maths Olympiad Prep

Library / /6 of 10

Number theory Difficulty 8.8 Shortlist Prove it United States

Find all pairs (a,n)(a, n) of positive integers with n>1n > 1 such that for every prime pp dividing an1a^n - 1, there exists 0<k<n0 < k < n such that pp divides ak1a^k - 1.

Solution

Define a good pair to be a pair (a,n)(a, n) of positive integers, with n>1n > 1, such that any prime dividing an1a^n - 1 also divides ak1a^k - 1 for some 0<k<n0 < k < n. We claim that the only good pairs are (2,6)(2, 6), (1,n)(1, n) for any nn, and (2m1,2)(2^m - 1, 2) for any m2m \ge 2.

It is straightforward to verify that all of these work, so it remains to show that these are the only good pairs. First, note that we have taken care of all cases where a=1a = 1, so we may henceforth assume a2a \ge 2. Also, note that if n=2n = 2, then (a,n)(a, n) is a good pair if and only if any prime dividing a21=(a+1)(a1)a^2 - 1 = (a+1)(a-1) also divides a1a - 1. Since the greatest common factor of a+1a+1 and a1a-1 is at most 22, it follows that a+1a+1 must be a power of 22. Thus, we have identified all the good pairs (a,n)(a, n) with n=2n = 2.

It remains to show that the only good pair (a,n)(a, n) with a2a \ge 2 and n3n \ge 3 is (2,6)(2, 6). Let Φn(x)\Phi_n(x) be the nthn^{\text{th}} cyclotomic polynomial, and for a prime pp, let vp(x)v_p(x) be the largest kk such that pkxp^k \mid x. Let ordp(x)\text{ord}_p(x) denote the order of xx modulo pp, meaning the smallest oo such that xo1(modp)x^o \equiv 1 \pmod p. We first prove a sequence of lemmas.

Lemma 4. Let n,kn, k, and dd be positive integers with n=kdn = kd. Then if aa is a positive integer and pp a prime such that either
* pp is odd and pak1p \mid a^k - 1, or
* p=2p = 2 and 4ak14 \mid a^k - 1,
then we have
vp(an1)=vp(ak1)+vp(d). v_p(a^n - 1) = v_p(a^k - 1) + v_p(d).
Proof. Consider the binomial expansion
an1=((ak1)+1)d1=i=1d(di)(ak1)i. a^n - 1 = ((a^k - 1) + 1)^d - 1 = \sum_{i=1}^{d} \binom{d}{i} (a^k - 1)^i.
Under either of the given assumptions, vp((di)(ak1)i)v_p\left(\binom{d}{i}(a^k - 1)^i\right) has a unique minimum at i=1i = 1, hence
vp(an1)=vp(d(ak1))=vp(ak1)+vp(d). v_p(a^n - 1) = v_p\left(d(a^k - 1)\right) = v_p(a^k - 1) + v_p(d). \quad \square

Lemma 5. For any n3n \ge 3 and any prime pp such that ordp(a)n\text{ord}_p(a) \ne n, we have vp(Φn(a))vp(n)v_p(\Phi_n(a)) \le v_p(n).
Proof. The statement obviously holds if vp(Φn(a))=0v_p(\Phi_n(a)) = 0. Let k=ordp(a)k = \text{ord}_p(a). Since Φn(x)xn1\Phi_n(x) \mid x^n - 1, it follows that knk \mid n.
Suppose pp is odd, and note that pak1p \mid a^k - 1. Lemma 4 applies, giving vp(an1)=vp(ak1)+vp(n/k)v_p(a^n - 1) = v_p(a^k - 1) + v_p(n/k). Since Φn(x)xn1xk1\Phi_n(x) \mid \frac{x^n-1}{x^k-1}, it follows that
vp(Φn(a))vp(an1ak1)=vp(n/k)vp(n). v_p(\Phi_n(a)) \le v_p\left(\frac{a^n - 1}{a^k - 1}\right) = v_p(n/k) \le v_p(n).
Next, we consider the special case when p=2p = 2. Since 2ak12 \mid a^k - 1, aa must be odd. Then, since Φn(x)xn1xk1\Phi_n(x) \mid \frac{x^n-1}{x^k-1}, we have
n1+a++an1an1a10(mod2). n \equiv 1 + a + \dots + a^{n-1} \equiv \frac{a^n - 1}{a - 1} \equiv 0 \pmod{2}.
Thus, nn is even, and we can write n=2mn = 2m.
Now, note that 4a214 \mid a^2 - 1. Also, since n>2n > 2, 22 is a proper divisor of nn, so Φn(x)xn1x21\Phi_n(x) \mid \frac{x^n-1}{x^2-1}. Thus, by Lemma 4 again, we have the remaining case
v2(Φn(a))v2(a2m1a21)=v2(m)v2(n). v_2(\Phi_n(a)) \le v_2\left(\frac{a^{2m} - 1}{a^2 - 1}\right) = v_2(m) \le v_2(n). \quad \square

Lemma 6. For any a2a \ge 2 and n>1n > 1, we have Φn(a)14aϕ(n)\Phi_n(a) \ge \frac{1}{4}a^{\phi(n)}.
Proof. Let pn=i=1n(12i)p_n = \prod_{i=1}^n (1 - 2^{-i}). We first prove that pn14p_n \ge \frac{1}{4} for all n1n \ge 1. It is clear that pn12p_n \le \frac{1}{2} for all n1n \ge 1. Hence, for each kk, pkpk+1=2k1pk2k2p_k - p_{k+1} = 2^{-k-1}p_k \le 2^{-k-2}. We therefore have
p1pn=i=1n1(pipi+1)i=1n12i214. p_1 - p_n = \sum_{i=1}^{n-1} (p_i - p_{i+1}) \le \sum_{i=1}^{n-1} 2^{-i-2} \le \frac{1}{4}.
Since p1=12p_1 = \frac{1}{2}, it follows that pn14p_n \ge \frac{1}{4}. We can now bound Φn(a)\Phi_n(a). We have
Φn(a)=dn(an/d1)μ(d)=adn(n/d)μ(d)dn(1an/d)μ(d)aϕ(n)i=1n(1ai)aϕ(n)i=1n(12i)14aϕ(n). \begin{align*} \Phi_n(a) &= \prod_{d|n} (a^{n/d} - 1)^{\mu(d)} = a^{\sum_{d|n} (n/d)\mu(d)} \prod_{d|n} (1 - a^{-n/d})^{\mu(d)} \\ &\geq a^{\phi(n)} \prod_{i=1}^{n} (1 - a^{-i}) \geq a^{\phi(n)} \prod_{i=1}^{n} (1 - 2^{-i}) \geq \frac{1}{4} a^{\phi(n)}. \quad \Box \end{align*}

Lemma 7. The only solutions to aϕ(n)4na^{\phi(n)} \le 4n, where a2a \ge 2 and n3n \ge 3 are integers, are (a,n)=(4,4),(3,3),(3,4)(a, n) = (4, 4), (3, 3), (3, 4), and (2,x)(2, x) where x{3,4,5,6,8,12,24}x \in \{3, 4, 5, 6, 8, 12, 24\}.
Proof. Let f(m)=2ϕ(m)mf(m) = \frac{2^{\phi(m)}}{m}, and note that ff is multiplicative. For any prime pp and any k2k \ge 2,
f(pk)=2pk1(p1)pk=2pk2(p1)2pf(pk1)2(p1)2pf(pk1)f(pk1), f(p^k) = \frac{2^{p^{k-1}(p-1)}}{p^k} = \frac{2^{p^{k-2}(p-1)^2}}{p} f(p^{k-1}) \ge \frac{2^{(p-1)^2}}{p} f(p^{k-1}) \ge f(p^{k-1}),
where we have used the fact that 2(p1)2p2^{(p-1)^2} \ge p when p2p \ge 2. It is easy to check that f(p)>4f(p) > 4 for any prime p7p \ge 7. Furthermore, f(n)>4f(n) > 4 for n=24,32,52n = 2^4, 3^2, 5^2. Thus, if f(n)4f(n) \le 4, then nn must be a divisor of 23352^3 \cdot 3 \cdot 5. Noting that f(2)=1,f(4)=1,f(8)=2,f(3)=43f(2) = 1, f(4) = 1, f(8) = 2, f(3) = \frac{4}{3}, and f(5)=165f(5) = \frac{16}{5}. From this and the multiplicativity of ff, it is easy to see that the only solutions of f(n)4f(n) \le 4 are n=2,3,4,5,6,8,12,24n = 2, 3, 4, 5, 6, 8, 12, 24. This shows that n{3,4,5,6,8,12,24}n \in \{3, 4, 5, 6, 8, 12, 24\} are the only solutions for a=2a = 2 with n3n \ge 3. If a3a \ge 3 and ϕ(n)4\phi(n) \ge 4, then aϕ(n)n=f(n)(a/2)ϕ(n)>4\frac{a^{\phi(n)}}{n} = f(n) \cdot (a/2)^{\phi(n)} > 4. Thus, the only possible solutions for a3a \ge 3 must satisfy n{3,4}n \in \{3, 4\}. Checking, we see that the only other solutions are (a,n)=(4,4),(3,3),(3,4)(a, n) = (4, 4), (3, 3), (3, 4). \square

We can now find all good pairs (a,n)(a, n) with a2a \ge 2 and n3n \ge 3. First, note that if (a,n)(a, n) is a good pair, then for any prime pan1p \mid a^n - 1, we must not have ordp(a)=n\text{ord}_p(a) = n. Hence, by Lemma 5, we must have vp(Φn(a))vp(n)v_p(\Phi_n(a)) \le v_p(n). This applies in particular to all primes dividing Φn(a)\Phi_n(a), so Φn(a)n\Phi_n(a) \le n.
By Lemma 6, we must then have naϕ(n)4n \ge \frac{a^{\phi(n)}}{4}. Thus, we see that the pairs in Lemma 7 are the only possible good pairs. Checking these finitely many cases, we find that (2,6)(2, 6) is the only good pair with a2a \ge 2 and n3n \ge 3, as needed.

Want a route through all this instead of an archive? The track puts 2,000 problems in a working order, from AMC 10 level to the IMO shortlist.

Source: MathNet, licensed CC-BY-4.0. Statement reproduced verbatim; metadata (topic, difficulty) added by this project.